[navinsylvester]

Interesting read.

But i bumped into the same problem recently while working on distributed simulation. I was writing an agent for an application which talks to clients over ssh. I started getting the same error when i cranked up machine count to 60.

I googled like every layman and zeroed in on MaxStartups. Looked at the manual to cross verify and fixed the issue in 5 minutes. I agree with what ever he did after he found out the issue but it was a clear case of SSH throwing an error. Why not just look up manual or google instead of wasting time on laborious packet inspection.

[toast0]

> Why not just look up manual or google instead of wasting time on laborious packet inspection.

When you've got a load balancing layer, and virtual machines, and interacting with the real world, there's a reasonable chance that the problem is not a (relatively) simple application configuration issue. If you've looked through enough of these, getting a pcap (especially with an identified customer) and looking through it with Wireshark is relatively quick.

[navinsylvester]

I could have worded it better like "why packet inspection to start with".

[sethammons]

I love packet inspection, but I always feel that if I have to go to that depth, then we missed an opportunity to have observability (metrics/logging) earlier.

[toast0]

Because pcap lies a lot less than everything else.

[abrookewood]

Because at their scale, people actually need to understand the issue rather than just applying fixes blindly plucked from the Internet. Don't get me wrong, Google is my go-to as well, but I suspect that they need to dig a little deeper.

[navinsylvester]

Fair enough.

But wouldn't doing investigation on the error provide data point to help debug the issue. They are trying to fix a customer issue and time is also of importance here.

> just applying fixes blindly plucked from the Internet.

I wasn't recommending that.